a survey of energy efficient network protocols forwireless networks
TRANSCRIPT
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
1/16
Wireless Networks 7, 343358, 2001
2001 Kluwer Academic Publishers. Manufactured in The Netherlands.
A Survey of Energy Efficient Network Protocols for Wireless
Networks
CHRISTINE E. JONES
BBN Technologies, Cambridge, MA 02138, USA
KRISHNA M. SIVALINGAM
School of EECS, Washington State University, Pullman, WA 99164, USA
PRATHIMA AGRAWAL and JYH CHENG CHENTelcordia Technologies, Morristown, NJ 07960, USA
Abstract. Wireless networking has witnessed an explosion of interest from consumers in recent years for its applications in mobile and
personal communications. As wireless networks become an integral component of the modern communication infrastructure, energy
efficiency will be an important design consideration due to the limited battery life of mobile terminals. Power conservation techniques
are commonly used in the hardware design of such systems. Since the network interface is a significant consumer of power, considerable
research has been devoted to low-power design of the entire network protocol stack of wireless networks in an effort to enhance energy
efficiency. This paper presents a comprehensive summary of recent work addressing energy efficient and low-power design within all
layers of the wireless network protocol stack.
Keywords: wireless networks, mobile computing, energy efficient design, network protocols, power aware protocols, low-power design
1. Introduction
The rapid expansion of wireless services such as cellular
voice, PCS (Personal Communications Services), mobiledata and wireless LANs in recent years is an indication that
significant value is placed on accessibility and portability as
key features of telecommunication [50]. Wireless devices
have maximum utility when they can be used anywhere
at anytime. One of the greatest limitations to that goal,
however, is finite power supplies. Since batteries provide
limited power, a general constraint of wireless communica-
tion is the short continuous operation time of mobile termi-
nals. Therefore, power management is one of the most chal-
lenging problems in wireless communication, and recent re-
search has addressed this topic [7]. Examples include a col-
lection of papers available in [72] and a recent conferencetutorial [54], both devoted to energy efficient design of wire-
less networks.
Studies show that the significant consumers of power in
a typical laptop are the microprocessor (CPU), liquid crystal
display (LCD), hard disk, system memory (DRAM), key-
board/mouse, CDROM drive, floppy drive, I/O subsystem,
and the wireless network interface card [55,62]. A typical
example from a Toshiba 410 CDT mobile computer demon-
Part of the research was supported by Air Force Office of Scientific
Research grants F-49620-97-1-0471 and F-49620-99-1-0125; by Tel-
cordia Technologies and by Intel. Part of the work was done while the first author was at Washington
State University. Corresponding author.
strates that nearly 36% of power consumed is by the display,
21% by the CPU/memory, 18% by the wireless interface,
and 18% by the hard drive. Consequently, energy conserva-
tion has been largely considered in the hardware design of
the mobile terminal [10] and in components such as CPU,
disks, displays, etc. Significant additional power savings
may result by incorporating low-power strategies into the
design of network protocols used for data communication.
This paper addresses the incorporation of energy conserva-
tion at all layers of the protocol stack for wireless networks.
The remainder of this paper is organized as follows. sec-
tion 2 introduces the network architectures and wireless pro-
tocol stack considered in this paper. Low-power design
within the physical layer is briefly discussed in section 2.3.
Sources of power consumption within mobile terminals and
general guidelines for reducing the power consumed are pre-
sented in section 3. Section 4 describes work dealing with
energy efficient protocols within the MAC layer of wire-
less networks, and power conserving protocols within the
LLC layer are addressed in section 5. Section 6 discusses
power aware protocols within the network layer. Opportu-
nities for saving battery power within the transport layer are
discussed in section 7. Section 8 presents techniques at the
OS/middleware and application layers for energy efficient
operation. Finally, section 9 summarizes and concludes thepaper.
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
2/16
344 JONES ET AL.
Figure 1. Infrastructure wireless network architecture.
2. Background
This section describes the wireless network architectures
considered in this paper. Also, a discussion of the wireless
protocol stack is included along with a brief description of
each individual protocol layer. The physical layer is further
discussed.
2.1. Wireless network architectures
Two different wireless network architectures are considered
in this paper: infrastructure and ad hoc networks. Below, a
description of each system architecture is presented.
Infrastructure. Wireless networks often extend, rather than
replace, wired networks, and are referred to as infrastruc-
ture networks. The infrastructure network architecture is de-
picted in figure 1. A hierarchy of wide area and local area
wired networks is used as the backbone network. The wired
backbone connects to special switching nodes called base
stations. Base stations are often conventional PCs and work-
stations equipped with custom wireless adapter cards. They
are responsible for coordinating access to one or more trans-
mission channel(s) for mobiles located within the coverage
cell. Transmission channels may be individual frequencies
in FDMA (Frequency Division Multiple Access), time slots
in TDMA (Time Division Multiple Access), or orthogonal
codes or hopping patterns in the case of CDMA (Code Divi-
sion Multiple Access). Therefore, within infrastructure net-
works, wireless access to and from the wired host occurs
in the last hop between base stations and mobile hosts that
share the bandwidth of the wireless channel.
Ad hoc. Ad hoc networks, on the other hand, are multi-
hop wireless networks in which a set of mobiles coopera-
tively maintain network connectivity [39]. This on-demand
network architecture is completely un-tethered from phys-
ical wires. An example of an ad hoc topology is picturedin figure 2. Ad hoc networks are characterized by dynamic,
Figure 2. Ad hoc wireless network architecture.
Figure 3. Protocol stack of a generic wireless network, and corresponding
areas of energy efficient research.
unpredictable, random, multi-hop topologies with typically
no infrastructure support. The mobiles must periodically ex-
change topology information which is used for routing up-
dates. Ad hoc networks are helpful in situations in which
temporary network connectivity is needed, and are oftenused for military environments, disaster relief, and so on.
Mobile ad hoc networks have attracted considerable atten-
tion as evidenced by the IETF working group MANET (Mo-
bile Ad hoc Networks). This has produced various Internet
drafts, RFCs, and other publications [38,39]. Also, a recent
conference tutorial presents a good introduction to ad hoc
networks [63].
2.2. Protocol layers
This section provides an introduction to the software used in
wireless data network systems. Application programs us-
ing the network do not interact directly with the networkhardware. Instead, an application interacts with the protocol
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
3/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 345
software. The notion of protocol layering provides a con-
ceptual basis for understanding how a complex set of proto-
cols work together with the hardware to provide a powerful
communication system. Recently, communication protocol
stacks such as the Infrared Data Association (IrDa) proto-
col stack for point-to-point wireless infrared communicationand the Wireless Application Protocol (WAP) Forum proto-
col stack for enabling developers to build advanced services
across differing wireless network technologies [25,64] have
been developed specifically for wireless networks. This pa-
per focuses on the traditional OSI protocol stack, depicted
in figure 3, for a generic wireless communication system.
The application and services layer occupies the top of the
stack followed by the operating system/middleware, trans-
port, network, data link, and physical layers. The problems
inherent to the wireless channel and issues related to mo-
bility challenge the design of the protocol stack adopted for
wireless networks. In addition, networking protocols need
to be designed with energy efficiency in mind.
Physical. The physical layer consists of radio frequency
(RF) circuits, modulation, and channel coding systems.
From an energy efficient perspective, considerable attention
has already been given to the design of this layer [10].
Data link. The data link layer is responsible for establish-
ing a reliable and secure logical link over the unreliable wire-
less link. The data link layer is thus responsible for wireless
link error control, security (encryption/decryption), mapping
network layer packets into frames, and packet retransmis-
sion.
A sublayer of the data link layer, the media access con-
trol (MAC) protocol layer is responsible for allocating the
time-frequency or code space among mobiles sharing wire-
less channels in a region.
Network. The network layer is responsible for routing
packets, establishing the network service type (connection-
less versus connection-oriented), and transferring packets
between the transport and link layers. In a mobile envi-
ronment this layer has the added responsibility of rerouting
packets and mobility management.
Transport. The transport layer is responsible for providingefficient and reliable data transport between network end-
points independent of the physical network(s) in use.
OS/Middleware. The operating system and middleware
layer handles disconnection, adaptivity support, and power
and quality of service (QoS) management within wireless
devices. This is in addition to the conventional tasks such as
process scheduling and file system management.
Application. The application and services layer deals with
partitioning of tasks between fixed and mobile hosts, source
coding, digital signal processing, and context adaptation in
a mobile environment. Services provided at this layer arevaried and application specific.
The next section further examines the low-power research
completed at the physical layer.
2.3. Physical layer
In the past, energy efficient and low-power design researchhas centered around the physical layer due to the fact that
the consumption of power in a mobile computer is a direct
result of the system hardware. Research addresses two dif-
ferent perspectives of the energy problem: (i) an increase in
battery capacity, and (ii) a decrease in the amount of energy
consumed at the wireless terminal.
The primary problem concerning energy in wireless com-
puting is that battery capacity is extremely limited. The fo-
cus of battery technology research has been to increase bat-
tery power capacity while restricting the weight of the bat-
tery. However, unlike other areas of computer technology
such as micro-chip design, battery technology has not expe-
rienced significant advancement in the past 30 years. There-
fore, unless a breakthrough occurs in battery technology, an
attainable goal of research would be a decrease in the energy
consumed in the wireless terminal [33].
Low-power design at the hardware layer uses different
techniques including variable clock speed CPUs [22], flash
memory [41], and disk spindown [17]. Numerous energy ef-
ficient techniquesfor the physical layer are discussed in [10].
Although the above techniques have resulted in considerable
energy savings, other venues should also be explored to im-
prove energy efficiency. One way to achieve this for future
wireless networks is to design the higher layers of the proto-
col stack with energy efficiency as an important goal.
3. Power consumption sources and conservation
mechanisms
This section first presents the chief sources of power con-
sumption with respect to the protocol stack. Then, it presents
an overview of the main mechanisms and principles that may
be used to develop energy efficient network protocols.
3.1. Sources of power consumption
The sources of power consumption, with regard to network
operations, can be classified into two types: communication
related and computation related.
Communication involves usage of the transceiver at the
source, intermediate (in the case of ad hoc networks), and
destination nodes. The transmitter is used for sending con-
trol, route request and response, as well as data packets orig-
inating at or routed through the transmitting node. The re-
ceiver is used to receive data and control packets some
of which are destined for the receiving node and some of
which are forwarded. Understanding the power characteris-
tics of the mobile radio used in wireless devices is important
for the efficient design of communication protocols. A typi-cal mobile radio may exist in three modes: transmit, receive,
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
4/16
346 JONES ET AL.
and standby. Maximum power is consumed in the transmit
mode, and the least in the standby mode. For example, the
Proxim RangeLAN2 2.4 GHz 1.6 Mbps PCMCIA card re-
quires 1.5 W in transmit, 0.75 W in receive, and 0.01 W
in standby mode. In addition, turnaround between transmit
and receive modes (and vice-versa) typically takes between6 and 30 microseconds. Power consumption for Lucents 15
dBm 2.4 GHz 2 Mbps Wavelan PCMCIA card is 1.82 W
in transmit mode, 1.80 W in receive mode, and 0.18 W in
standby mode. Thus, the goal of protocol development for
environmentswith limited power resources is to optimize the
transceiver usage for a given communication task.
The computation considered in this paper is chiefly con-
cerned with protocol processing aspects. It mainly involves
usage of the CPU and main memory and, to a very small ex-
tent, the disk or other components. Also, data compression
techniques, which reduce packet length (and hence energy
usage), may result in increased power consumption due to
increased computation.
There exists a potential tradeoff between computation
and communication costs. Techniques that strive to achieve
lower communication costs may result in higher computa-
tion needs, and vice-versa. Hence, protocols that are devel-
oped with energy efficiency goals should attempt to strike a
balance between the two costs.
3.2. General conservation guidelines and mechanisms
The following discussion presents some general guidelines
that may be adopted for an energy efficient protocol design,
and figure 3 lists areas in which conservation mechanismsare beneficial. Examples are provided in which these guide-
lines have been adopted. Some mechanisms are better suited
for infrastructure networks and others for ad hoc networks.
Collisions should be eliminated as much as possible
within the MAC layer since they result in retransmissions.
Retransmissions lead to unnecessary power consumption
and to possibly unbounded delays. Retransmissions cannot
be completely avoided in a wireless network due to the high
error-rates. Similarly, it may not be possible to fully elimi-
nate collisions in a wireless mobile network. This is partly
due to user mobility and a constantly varying set of mobiles
in a cell. For example, new users registering with the base
station may have to use some form of random access pro-
tocol. In this case, using a small packet size for registra-
tion and bandwidth request may reduce energy consumption.
The EC-MAC protocol [53] is one example that avoids col-
lisions during reservation and data packet transmission.
In a typical broadcast environment, the receiver remains
on at all times which results in significant power consump-
tion. The mobile radio receives all packets, and forwards
only the packets destined for the receiving mobile. This is
the default mechanism used in the IEEE 802.11 wireless pro-
tocol in which the receiver is expected to keep track of chan-
nel status through constant monitoring. One solution is to
broadcast a schedule that contains data transmission startingtimes for each mobile as in [53]. This enables the mobiles to
switch to standby mode until the receive start time. Another
solution is to turn off the transceiver whenever the node de-
termines that it will not be receiving data for a period of time.
The PAMAS protocol [51] uses such a method.
Furthermore, significant time and power is spent by the
mobile radio in switching from transmit to receive modes,and vice versa. A protocol that allocates permission on a
slot-by-slot basis suffers substantial overhead. Therefore,
this turnaround is a crucial factor in the performance of a
protocol. If possible, the mobile should be allocated con-
tiguous slots for transmission or reception to reduce turn-
around, resulting in lower power consumption. This is simi-
lar to buffering writes to the hard disk in order to minimize
seek latency and head movement. Also, it is beneficial for
mobiles to request multiple transmission slots with a sin-
gle reservation packet when requesting bandwidth in order
to reduce the reservation overhead. This leads to improved
bandwidth usage and energy efficiency. The scheduling al-
gorithms studied in [13] consider contiguous allocation and
aggregate packet requests.
Assuming that mobiles transmit data transmission re-
quests to the base station, a centralized scheduling mecha-
nism that computes the system transmission schedule at the
base station is more energy efficient. A distributed algorithm
in which each mobile computes the schedule independently
may not be desirable because mobiles may not receive all
reservation requests due to radio and error constraints, and
schedule computation consumes energy. Thus, computation
of the transmission schedule ought to be relegated to the base
station, which in turn broadcasts the schedule to each mo-
bile. Most reservation and scheduling based protocols re-quire the base station to compute the schedule.
The scheduling algorithm at the base station may con-
sider the nodes battery power level in addition to the con-
nection priority. This allows traffic from low-power mobiles
that may be dropped due to depletion of power reserves to
be transmitted sooner. Such a mechanism has been stud-
ied in [47,48]. Also, under low-power conditions, it may be
useful to allow a mobile to re-arrange allocated slots among
its own flows. This may allow certain high-priority traffic
to be transmitted sooner rather than waiting for the orig-
inally scheduled time. Such mobile-based adaptive algo-
rithms have been considered in [14,15] in the context of en-
ergy efficiency and channel error compensation.
At the link layer, transmissions may be avoided when
channel conditions are poor, as studied in [69]. Also, er-
ror control schemes that combine automatic repeat request
(ARQ) and forward error correction (FEC) mechanisms may
be used to conserve power (i.e., tradeoff retransmissions
with ARQ versus longer packets with FEC) as in [32].
Energy efficient routing protocols may be achieved by es-
tablishing routes that ensure that all nodes equally deplete
their battery power, as studied in [11,68]. This helps balance
the amount of traffic carried by each node. A related mech-
anism is to avoid routing through nodes with lower battery
power, but this requires a mechanism for dissemination ofnode battery power. Also, the periodicity of routing updates
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
5/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 347
can be reduced to conserve energy, but may result in inef-
ficient routes when user mobility is high. Another method
to improve energy performance is to take advantage of the
broadcast nature of the network for broadcast and multicast
traffic as in [52,66]. In [49], the topology of the network is
controlled by varying the transmit power of the nodes, andthe topology is generated to satisfy certain network proper-
ties.
At the OS level, the common factor to all the different
techniques proposed is suspension of a specific sub-unit such
as disk, memory, display, etc. based upon detection of pro-
longed inactivity. Several methods of extending battery life-
time within the operating system and middleware layer are
discussed in [10,42,58]. Other techniques studied include
power-aware CPU scheduling [36,65] and page allocation
[31]. Within the application layer, the power conserving
mechanisms tend to be application specific such as data-
base access [3,24] and video processing [1,10,21]. A sum-
mary of software strategies for energy efficiency is presentedin [37].
4. MAC sublayer
The MAC (Media Access Control) layer is a sublayer of the
data link layer which is responsible for providing reliability
to upper layers for the point-to-pointconnections established
by the physical layer. The MAC sublayer interfaces with
the physical layer and is represented by protocols that define
how the shared wireless channels are to be allocated among
a number of mobiles. This section presents the details of
three specific MAC protocols: IEEE 802.11 [23], EC-MAC[53], and PAMAS [51].
4.1. IEEE 802.11 standard
The IEEE 802.11 [23] protocol for wireless LANs is a mul-
tiple access technique based on CSMA/CA (Collision Sense
Multiple Access/Collision Avoidance), and is derived from
the MACA protocol described in [29]. The basic protocol
is defined as follows. A mobile with a packet to transmit
senses the transmission channel for activity. The mobile
captures the channel and transmits all pending data pack-
ets if the channel is not busy. Otherwise, the mobile defers
transmission and enters the backoff state. The time period
that follows is called the contention window and consists of
a pre-determined number of transmission slots. The mobile
randomly selects a slot in the contention window, and contin-
uously senses the medium until its selected contention slot.
The mobile enters the backoff state again if it detects trans-
mission from some other mobile during that period. How-
ever, if no transmission is detected, the mobile transmits the
access packet and captures the channel. Extensions to the
basic protocol include provisions for MAC-level acknowl-
edgements and request-to-send (RTS)/clear-to-send (CTS)
mechanisms.
The IEEE 802.11 [23] standard recommends the follow-ing technique for power conservation. A mobile that wishes
to conserve power may switch to sleep mode and inform
the base station of this decision. The base station buffers
packets received from the network that are destined for the
sleeping mobile. The base station periodically transmits a
beacon that contains information about such buffered pack-
ets. When the mobile wakes up, it listens for this beacon,and responds to the base station which then forwards the
packets. This approach conserves power but results in ad-
ditional delays at the mobile that may affect the quality of
service (QoS). A comparison of power-saving mechanisms
in the IEEE 802.11 and HIPERLAN standards is presented
in [67]. Presented in [16] is a load-sharing method for sav-
ing energy in an IEEE 802.11 network. Simulation results
indicate total power savings of 515%.
Experimental measurements of per-packet energy con-
sumption for an IEEE 802.11 wireless network interface is
reported in [19]. This work uses the Lucent WaveLAN card
for its experiments. The cost of sending and receiving a
packet is measured for a network using UDP point-to-point
(or unicast) and broadcast traffic with varying packet sizes.
The energy cost is studied in terms of fixed cost per packet
which reflects MAC operation and incremental cost that de-
pends on packet size. The results show that both point-to-
point and broadcast traffic transmission incur the same incre-
mental costs, but point-to-point transmission incurs higher
fixed costs because of the MAC coordination. The reception
of point-to-point traffic maintains higher fixed costs since
the receiver must respond with CTS and ACK messages.
However, incremental costs of packet reception were iden-
tical for both traffic types. The study also measures power
consumption for non-destination mobiles that are in rangeof the sender and receiver. These experiments are a valu-
able source of information and represent an important step
in expanding the knowledge of energy efficient protocol de-
velopment.
4.2. EC-MAC protocol
Although the IEEE 802.11 standard addresses energy effi-
ciency, it was not one of the central design issues in de-
veloping the protocol. The EC-MAC (Energy Conserving-
Medium Access Control) protocol [12,53], on the other
hand, was developed with the issue of energy efficiency as a
primary design goal. The EC-MAC protocol is defined for
an infrastructure network with a single base station serving
mobiles in its coverage area. This definition can be extended
to an ad hoc network by allowing the mobiles to elect a co-
ordinator to perform the functions of the base station. The
general guidelines outlined in the previous section and the
need to support QoS led to a protocol that is based on reser-
vation and scheduling strategies. Transmission in EC-MAC
is organized by the base station into frames as shown in fig-
ure 4, and each slot equals the basic unit of wireless data
transmission.
At the start of each frame, the base station transmits
the frame synchronization message (FSM) which containssynchronization information and the uplink transmission or-
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
6/16
348 JONES ET AL.
Figure 4. EC-MAC protocol frame structure.
der for the subsequent reservation phase. During the re-
quest/update phase, each registered mobile transmits new
connection requests and status of established queues accord-
ing to the transmission order received in the FSM. In this
phase, collisions are avoided by having the BS send the ex-
plicit order of reservation transmission. New mobiles that
have entered the cell coverage area register with the base sta-
tion during the new-user phase. Here, collisions are not eas-
ily avoided and hence this may be operated using a variant
of Aloha. This phase also provides time for the BS to com-
pute the data phase transmission schedule. The base station
broadcasts a schedule message that contains the slot permis-
sions for the subsequent data phase. Downlink transmission
from the base station to the mobile is scheduled considering
the QoS requirements. Likewise, the uplink slots are allo-cated using a suitable scheduling algorithm.
Energy consumption is reduced in EC-MAC because of
the use of a centralized scheduler. Therefore, collisions over
the wireless channel are avoided and this reduces the num-
ber of retransmissions. Additionally, mobile receivers are
not required to monitor the transmission channel as a result
of communication schedules. The centralized scheduler may
also optimize the transmission schedule so that individual
mobiles transmit and receive within contiguous transmission
slots. The priority round robin with dynamic reservation up-
date and error compensation scheduling algorithm described
in [13] provides for contiguous slot allocation in order to
reduce transceiver turnaround. Also, scheduling algorithms
that consider mobile battery power level in addition to packet
priority may improve performance for low-power mobiles.
A family of algorithms based on this idea is presented in
[30,48].
The frames may be designed to be fixed or variable
length. Fixed length frames are desirable from the energy ef-
ficiency perspective, since a mobile that goes to sleep mode
will know when to wake up to receive the FSM. However,
variable length frames are better for meeting the demands
of bursty traffic. The EC-MAC studies used fixed length
frames.
The energy efficiency of EC-MAC is compared with thatof IEEE 802.11 and other MAC protocols in [12]. This
comparative study demonstrates how careful reservation and
scheduling of transmissions avoids collisions that are expen-
sive in energy consumption.
4.3. PAMAS protocol
While the EC-MAC protocol described above was designed
primarily for infrastructure networks, the PAMAS (Power
Aware Multi-Access) protocol [51] was designed for the ad
hoc network, with energy efficiency as the primary design
goal.
The PAMAS protocol modifies the MACA protocol de-
scribed in [29] by providing separate channels for RTS/CTS
control packets and data packets. In PAMAS, a mobile witha packet to transmit sends a RTS (request-to-send) message
over the control channel, and awaits the CTS (clear-to-send)
reply message from the receiving mobile. The mobile en-
ters a backoff state if no CTS arrives. However, if a CTS is
received, then the mobile transmits the packet over the data
channel. The receiving mobile transmits a busy tone over
the control channel enabling users tuned to the control chan-
nel to determine that the data channel is busy.
Power conservation is achieved by requiring mobiles that
are not able to receive and send packets to turn off the wire-
less interface. The idea is that a data transmission between
two mobiles need not be overheard by all the neighbors ofthe transmitter. The use of a separate control channel al-
lows for mobiles to determine when and for how long to
power off. A mobile should power itself off when: (i) it
has no packets to transmit and a neighbor begins transmit-
ting a packet not destined for it, and (ii) it does have pack-
ets to transmit but at least one neighbor-pair is communicat-
ing. Each mobile determines the length of time that it should
be powered off through the use of a probe protocol, the de-
tails of which are available in [51]. Theoretical bounds on
power savings for random, line, and fully connected topolo-
gies are also presented. The results from simulation and
analysis show that between 10% and 70% power savings canbe achieved for fully connected topologies.
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
7/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 349
5. LLC sublayer
In this section, we focus on the error control functionality of
the logical link control (LLC) sublayer. The two most com-
mon techniques used for error control are Automatic Repeat
Request (ARQ) and Forward Error Correction (FEC). Both
ARQ and FEC error control methods waste network band-
width and consume power resources due to retransmission
of data packets and greater overhead necessary in error cor-
rection. Care must be exercised while adopting these tech-
niques over a wireless link where the error rates are high
due to noise, fading signals, and disconnections caused from
mobility. A balance needs to be maintained within this layer
between competing measures for enhancing throughput, re-
liability, security, and energy efficiency. For example, chan-
nel encoding schemes for enhancing channel quality tend to
reduce the throughput as more redundancy is added to the
transmitted information. Also, increasing transmitted power
to improve the channel to interference ratio depletes batteryenergy.
Recent research has addressed low-power error control
and several energy efficient link layer protocols have been
proposed. Three such protocols are described below.
5.1. Adaptive error control with ARQ
An ARQ strategy that includes an adaptive error control pro-
tocol is presented and studied in [69,70]. First, though, the
authors propose a new design metric for protocols developed
specifically for the wireless environment and three guide-
lines in designing link layer protocols to be more power con-
serving. The new design metric introduced in [69] is the
energy efficiency of a protocol which is defined as the ra-
tio between total amount of data delivered and total energy
consumed. Therefore, as more data is successfully transmit-
ted for a given amount of energy consumption, the energy
efficiency of the protocol increases.
The following guidelines in developing a protocol should
be considered in order to maximize the energy efficiency of
the protocol.
1. Avoid persistence in retransmitting data.
2. Trade off number of retransmission attempts for proba-
bility of successful transmission.3. Inhibit transmission when channel conditions are poor.
The energy efficient protocol proposed in [69,70] incor-
porates a probing protocol that slows down data transmission
when degraded channel conditions are encountered. The
ARQ protocol works as normal until the transmitter detects
an error in either the data or control channel due to the lack
of a received acknowledgement (ACK). At this time the pro-
tocol enters a probing mode in which a probing packet is
transmitted every t slots. The probe packet contains only
a header with little or no payload and therefore consumes
a smaller amount of energy. This mode is continued until
a properly received ACK is encountered, indicating the re-covered status of both channels. The protocol then returns to
normal mode and continues data transmission from the point
at which it was interrupted.
Using a Markov model based analysis and a recursive
technique, the ARQ probing protocol is compared to tradi-
tional ARQ schemes, and the tradeoff between performance
and energy efficiency is investigated. The results show thatunder slow fading channel conditions the proposed protocol
is superior to that of standard ARQ in terms of energy ef-
ficiency, increasing the total number of packets that can be
transmitted. The analysis also demonstrates that an optimal
transmission power in respect to energy efficiency exists.
Using a high transmission power to maximize the probabil-
ity of a successful transmission may not be the best strategy.
Although decreasing the transmission power results in an in-
creased number of transmission attempts, it may be more ef-
ficient than attempting to maximize the throughput per slot.
The conclusion reached is that although throughput is not
necessarily maximized, the energy efficiency of a protocolmay be maximized by decreasing the number of transmis-
sion attempts and/or transmission power in the wireless en-
vironment.
5.2. Adaptive error control with ARQ/FEC combination
The above error control scheme included only ARQ strate-
gies. However, the energy efficient error control scheme pro-
posed by Lettieri et al. [32] combines ARQ and FEC strate-
gies. The authors describe an error control architecture for
the wireless link in which each packet stream maintains its
own time-adaptive customized error control scheme based
on certain set up parameters and a channel model estimated
at run-time. The idea behind this protocol is that there exists
no energy efficient one-size-fits-all error control scheme
for all traffic types and channel conditions. Therefore, er-
ror control schemes should be customized to traffic require-
ments and channel conditions in order to obtain more opti-
mal energy savings for each wireless connection.
The dynamic error control protocol described in [32] op-
erates as follows. Service quality parameters, such as packet
size and QoS requirements, used by the MAC sublayer and
packet scheduler are associated with each data stream. These
parameters are further used to select an appropriately cus-tomized combination of an ARQ scheme (Go-Back-N, Cu-
mulative Acknowledgement (CACK), Selective Acknowl-
edgement (SACK), etc.) and FEC scheme. In order to
keep energy consumption at a minimum, the error control
scheme associated with each stream may need to be modified
as channel conditions change over time. Studies based on
analysis and simulation under different scenarios were pre-
sented as a guideline in choosing an error control scheme to
achieve low energy consumption while trading off QoS over
various channel conditions, traffic types, and packet sizes.
The authors extend their research in [34] to include a proto-
col for dynamically sizing the MAC layer frame, dependingupon wireless channel conditions.
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
8/16
350 JONES ET AL.
5.3. Adaptive power control and coding scheme
Finally, a dynamic power control and coding protocol for op-
timizing throughput, channel quality, and battery life is stud-
ied in [2,44]. This distributed algorithm, in which each mo-
bile determines its own operating point with respect to powerand error control parameters, maintains the goal of minimiz-
ing power utilization and maximizing capacity in terms of
the number of simultaneous connections. Power control, as
defined by the authors, is the technique of controlling the
transmit power so as to affect receiver power, and ultimately
the carrier-to-interference ratio (CIR).
The energy efficient power control and coding protocol
operates in the following manner. Each transmitter operates
at a power-code pair in which the power level lies between
a specified minimum and maximum and the error code is
chosen from a finite set. The algorithm is iterative in nature
with the transmitter and receiver, at each iteration, cooper-
atively evaluating channel performance and determining if
an adjustment in the power-code pair is necessary. The time
between each iteration is referred to as a timeframe. Af-
ter each timeframe the receiver involved in the data trans-
mission evaluates the channel performance by checking the
word error rate (WER). If the WER lies within an accept-
able range, the power-code pair is retained; otherwise a new
power-code pair is computed by the transmitter. The basic
frame of the algorithm can be modified such that optimal
levels of control overhead and channel quality are traded off.
Also, variations of the base algorithm include the evaluation
of the average WER, rather than the instantaneous WER in
each timeframe, in determining channel quality and in theevaluation of anticipated channel performance. The latter
adaptation of the algorithm attempts to predict changes in
error rates due to mobility by sampling the received pow-
ers and extrapolating the values to the next timeframe. If
the predicted WER is not within acceptable ranges, then the
power-code pair is adapted to avoid unsatisfactory channel
conditions.
A study of the dynamic power control and coding proto-
col was performed through simulation of a cellular system
with roaming mobiles. Simulation results indicate that the
proposed dynamic power control and coding protocol sup-
ports better quality channels as compared to schemes that
use fixed codes; therefore power-control alone does not per-
form as well as an adaptive power-control/FEC protocol.
The next section discusses energy efficient routing proto-
cols within the network layer.
6. Network layer
The main functions of the network layer are routing pack-
ets and congestion control. In wireless mobile networks,
the network layer has the added functionality of routing un-
der mobility constraints and mobility management including
user location, update, etc. In this section, we present en-ergy efficient routing algorithms developed for wireless ad
Figure 5. Example ad hoc topology.
hoc networks. Energy efficient routing does not apply to in-
frastructure networks because all traffic is routed through the
base station.
As mentioned earlier, in ad hoc networks the mobiles
cooperate to maintain topology information and use multi-
hop packet routing. The problem of routing is complicated
due to user mobility resulting in frequently changed network
topologies. The rate of topology change depends on many
factors including user mobility speeds and terrain character-
istics. Typical routing algorithms for ad hoc networks con-
sider two different approaches:
1. Use frequent topology updates resulting in improved
routing, but increased update messages consume precious
bandwidth.
2. Use infrequent topology updates resulting in decreased
update messages, but inefficient routing and occasionally
missed packets results.
Typical metrics used to evaluate ad hoc routing protocols
are shortest-hop, shortest-delay, and locality stability (Woo
et al. [68]). However, these metrics may have a negative ef-
fect in wireless networks because they result in the overuse
of energy resources of a small set of mobiles, decreasingmobile and network life. For example, consider the wireless
network in figure 5. Using shortest-hop routing, traffic from
mobile A to mobile D will always be routed through mobile
E, which will drain the energy reserves of E faster. If mobile
Es battery becomes fully drained, then mobile F is discon-
nected from the network and communication to and from F
is no longer possible. By using a routing algorithm that takes
into account such issues, traffic from A to D may not always
be routed through E, but through mobiles B and C, extend-
ing network life. Consequently, it is essential to consider
routing algorithms from an energy efficient perspective, in
addition to traditional metrics. Such research is described inthe following paragraphs.
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
9/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 351
6.1. Unicast traffic
Unicast traffic is defined as traffic in which packets are des-
tined for a single receiver. In [68], routing of unicast traffic is
addressed with respect to battery power consumption. The
authors research focuses on designing protocols to reduceenergy consumption and to increase the life of each mobile,
increasing network life as well. To achieve this, five differ-
ent metrics are defined from which to study the performance
of power-aware routing protocols.
Energy consumed per packet. It is easy to observe that if
energy consumed per packet is minimized then the total en-
ergy consumed is also minimized. Under light loads, this
metric will most likely result in the shortest-hop path. As
network load increases, this is not necessarily the case be-
cause the metric will tend to route packets around areas of
congestion in the network.
Time to network partition. Given a network topology, a
minimal set of mobiles exist such that their removal will
cause the network to partition. Routes between the two par-
titions must go through one of the critical mobiles; there-
fore a routing algorithm should divide the work among these
mobiles in such a way that the mobiles drain their power at
equal rates.
Variance in power levels across mobiles. The idea behind
this metric is that all mobiles in a network operate at the
same priority level. In this way, all mobiles are equal and
no one mobile is penalized or privileged over any other.
This metric ensures that all mobiles in the network remainpowered-on togetherfor as long as possible.
Cost per packet. In order to maximize the life of all mo-
biles in the network, metrics other than energy consumed per
packet need to be used. When using these metrics, routes
should be created such that mobiles with depleted energy
reserves do not lie on many routes. Together, these metrics
become the cost of a packet, which needs to be minimized.
Maximum mobile cost. This metric attempts to minimize
the cost experienced by a mobile when routing a packet
through it. By minimizing the cost per mobile, significant
reductions in the maximum mobile cost result. Also, mo-
bile failure is delayed and variance in mobile power levels is
reduced due to this metric.
In order to conserve energy, the goal is to minimize all the
metrics except for the second which should be maximized.
As a result, a shortest-hop routing protocol may no longer
be applicable; rather, a shortest-cost routing protocol with
respect to the five energy efficiency metrics would be per-
tinent. For example, a cost function may be adapted to ac-
curately reflect a batterys remaining lifetime. The premise
behind this approach is that although packets may be routed
through longer paths, the paths contain mobiles that havegreater amounts of energy reserves. Also, energy can be
conserved by routing traffic through lightly loaded mobiles
because the energy expended in contention and retransmis-
sion is minimized.
The properties of the power-aware metrics and the ef-
fect of the metrics on end-to-end delay are studied in [68]
using simulation. A comparison of shortest-hop routingand the power-aware shortest-cost routing schemes was con-
ducted. The performance measures were delay, average cost
per packet, and average maximum node cost. Results show
that usage of power-aware metrics result in no extra delay
over the traditional shortest-hop metric. This is true because
congested paths are often avoided. However, there was sig-
nificant improvement in average cost per packet and average
maximum mobile cost in which the cost is in terms of the
energy efficient metrics defined above. The improvements
were substantial for large networks and heavily-loaded net-
works. Therefore, by adjusting routing parameters a more
energy efficient routing scheme may be utilized for wireless
networks.
The above approach to routing in wireless ad hoc net-
works requires, at the least, that every mobile have knowl-
edge of the locations of every other mobile and the links
between them. This creates significant communication over-
head and increased delay. Research completed in [56] ad-
dresses this issue by proposing localized routing algorithms
which depend only on information about the source location,
the location of neighbors, and location of the destination.
This information is collected through GPS receivers which
are included within each mobile. Therefore, excessive net-
work communication is not required which, the authors re-
port, more than makes up for the extra energy consumed bythe GPS units.
A new power-cost metric incorporating both a mobiles
lifetime and distance based power metrics is proposed, and
using the newly defined metric, three power-aware localized
routing algorithms are developed: power, cost, and power-
cost. The power algorithm attempts to minimize the to-
tal amount of power utilized when transmitting a packet,
whereas the cost algorithm avoids mobiles that maintain low
battery reserves in order to extend the network lifetime. Fi-
nally, the power-cost routing algorithm is a combination of
the two algorithms. Experiments validated the performance
of these routing algorithms.
6.2. Broadcast traffic
In this section broadcast traffic, which is defined as traffic
in which packets are destined for all mobiles in the system,
is considered. With a single transmission, a mobile is able
to broadcast a packet to all immediate neighbors. However,
each mobile needs to receive a packet only once. Interme-
diate mobiles are required to retransmit the packet. The key
idea in conserving energy is to allow each mobiles radio to
turn off after receiving a packet if its neighbors have already
received a copy of the packet. Addressed in [52] is the rout-ing of broadcast traffic in terms of power consumption.
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
10/16
352 JONES ET AL.
The broadcast technique used in traditional networks is a
simple flooding algorithm. This algorithm gathers no global
topology information, requiring little control overhead and
completes the broadcast with minimum number of hops.
However, the flooding algorithm is not suitable for wireless
networks because many intermediate nodes must retransmitpackets needlessly which leads to excessive power consump-
tion. Therefore, the authors of [52] propose that it is more
beneficial to spend some energy in gathering topology in-
formation in order to determine the most efficient broadcast
tree.
In order to increase mobile and network life, any broad-
cast algorithm used in the wireless environment should focus
on conserving energy and sharing the cost of routing among
all mobiles in the system. One way to conserve power is by
ensuring that a transmission reaches as many new nodes as
possible. A broadcast tree approach is presented in [52], in
which the tree is constructed starting from a source and ex-
panding to the neighbor that has the lowest cost per outgoing
degree, where the cost associated with each mobile increases
as the mobile consumes more power. Therefore, priority for
routing packets through the broadcast tree is given to nodes
that have consumed lower amounts of power and nodes that
have more neighbors which have not already received the
data transmission. Since mobile costs continuously change,
broadcast transmissions originating from the same source
may traverse different trees, as they are determined based
on current costs of nodes.
Simulations were conducted in order to study the perfor-
mance of the proposed power-aware broadcast protocol as
compared to traditional flooding in terms of energy savingsas well as delay. Averaged over a period of time, the power-
aware protocol demonstrates very little difference in broad-
cast delay. However, results indicate that savings in energy
consumption of 20% or better are possible using the power-
aware broadcast algorithm, with greater savings in larger
networks and networks with increased traffic loads.
The construction of energy efficient broadcast and multi-
cast trees for the wireless environment is also studied in [66].
The authors state that mobiles may experience greater en-
ergy conservation if routing decisions are combined with de-
cisions concerning transmission power levels. An algorithm
is presented for determining the minimum-energy source-
based tree for each broadcast/multicast session request. The
algorithm is based on the concept that there exists an optimal
point in the trade-off between reaching greater number of
mobiles in a single hop by using higher transmission power
versus reaching fewer mobiles but using lower power lev-
els. Performance results demonstrate that the combination
of routing and transmission power decisions provide greater
energy conservation. A similar idea concerning the incorpo-
ration of transmission power levels into routing algorithms
is also presented for unicast traffic in [11].
In [18], a simulation based comparison of energy con-
sumption for two ad hoc routing protocols Dynamic
Source Routing (DSR) [28] and Ad hoc On Demand Vec-tor routing (AODV) [45] protocols is presented. The analy-
sis considers the cost for sending and receiving traffic, for
dropped packets, and for routing overhead packets. User
mobility is modeled in the analysis. The observations in-
dicate that energy spent on receiving and discarding packets
can be significant. Also, the costs of flooding-based broad-
cast traffic and MAC control were seen to be significant. ForDSR, results show that the cost of source routing headers
was not very high, but operating the receiver in promiscuous
mode for caching and route response purposes resulted in
high power consumption. Results also indicate that since
AODV generates broadcast traffic more often, the energy
cost is high given that broadcast traffic consumes more en-
ergy. Refer to [18] for more detailed results.
The next section presents work related to improving
transport protocol performance in the wireless environment.
7. Transport layer
The transport layer provides a reliable end-to-end data de-
livery service to applications running at the end points of a
network. The most commonly used transport protocol for
wired networks, where underlying physical links are fairly
reliable and packet loss is random in nature, is the Transmis-
sion Control Protocol (TCP) [46]. However, due to inher-
ent wireless link properties, the performance of traditional
transport protocols such as TCP degrades significantly over a
wireless link. TCP and similar transport protocols resort to a
larger number of retransmissions and frequently invoke con-
gestion control measures, confusing wireless link errors and
loss due to handoff as channel congestion. This can signifi-cantly reduce throughput and introduce unacceptable delays
[9]. As stated earlier, increased retransmissions unnecessar-
ily consume battery energy and limited bandwidth.
Recently, various schemes have been proposed to allevi-
ate the effects of non congestion-related losses on TCP per-
formance over networks with wireless links. These schemes,
which attempt to reduce retransmissions, are classified into
three basic groups: (i) split connection protocols, (ii) link-
layer protocols, and (iii) end-to-end protocols.
Split connection protocols completely hide the wireless
link from the wired network by terminating the TCP con-
nections at the base station as shown in figure 6. This is
accomplished by splitting each TCP connection between the
source and destination into two separate connections at the
base station. The result is one TCP connection between the
wired network and the base station and a second TCP con-
nection between the base station and the mobile. The second
connection over the wireless link may use modified versions
of TCP that enhance performance over the wireless channel.
Examples of split connection protocols include Indirect-TCP
[5], Berkeley Snoop Module [6], and M-TCP [8].
Figure 7 depicts the link layer approach which attempts to
hide link related losses from the TCP source by using a com-
bination of local retransmissions and forward error correc-
tion over the wireless link. Local retransmissions use tech-niques that are tuned to the characteristics of the wireless
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
11/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 353
Figure 6. Split connection TCP approach.
Figure 7. Link layer TCP approach.
channel to provide significant increase in performance. One
example of a link layer protocol is the AIRMAIL protocol[4], which employs a combination of both FEC and ARQ
techniques for loss recovery.
Finally, end-to-end protocols include modified versions
of TCP that are more sensitive to the wireless environ-
ment. End-to-end protocols require that a TCP source han-
dle losses through the use of such mechanisms as selective
acknowledgements and explicit loss notification (ELN). Se-
lective acknowledgements allow the TCP source to recover
from multiple packet losses, and ELN mechanisms aid the
TCP source in distinguishing between congestion and other
forms of loss.
7.1. Energy consumption analysis of TCP
The procotols described previously generally achieve higher
throughput rates over the wireless channel than standard
TCP because the protocols are better able to adapt to the
dynamic mobile environment. However, the performance of
a particular protocol is largely dependent upon various fac-
tors such as mobility handling, amount of overhead costs
incurred, frequency and handling of disconnections, etc.
Therefore, performance and energy conservation may range
widely for these protocols depending upon both internal al-
gorithm and external environmental factors. Although the
above protocols, along with many others proposed in re-search, have addressed the unique needs of designing trans-
port protocols in the wireless environment which may or
may not lead to greater energy efficiency, they have not di-rectly addressed the idea of a low-power transport protocol.
The energy consumption of Tahoe, Reno, and New Reno
versions of TCP is analyzed in [71]. Energy consumption
is the main parameter studied with the objective of measur-
ing the effect of TCP transmission policies on energy per-
formance. The energy efficiency of a protocol is defined
as the average number of successful transmissions per en-
ergy unit, which can be computed as the average number
of successes per transmission attempt. A two-state Markov
packet error process is used in the performance evaluation
of a single transceiver running the various versions of TCP
on a dedicated wireless link. Results of the study demon-
strate that error correlation significantly affects the energy
performance of TCP and that congestion control algorithms
of TCP actually allow for greater energy savings by backing
off and waiting during error bursts. It is also seen that en-
ergy efficiency may be quite sensitive to the version of TCP
implemented and the choice of protocol parameters.
The same versions of TCP were studied in [60] in
terms of energy/throughput tradeoffs. Simulation results
show that no single TCP version is most appropriate within
wired/wireless heterogenous networks, and that the key to
balancing energy and throughput performance is through
the error control mechanism. Using these results, the au-
thors propose a modified version of TCP, referred to as TCP-Probing, in [59].
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
12/16
354 JONES ET AL.
In TCP-Probing, data transmission is suspended and a
probe cycle is initiated when a data segment is delayed or
lost, rather than immediately invoking congestion control.
A probe cycle consists of an exchange of probe segments be-
tween sender and receiver. Probe segments are implemented
as extensions to the TCP header and carry no payload. TheTCP sender monitors the network through the probe cy-
cle which terminates when two consecutive round-trip-times
(RTT) are successfully measured. The sender invokes stan-
dard TCP congestion control if persistent error conditions
are detected. However, if monitored conditions indicate tran-
sient random error, then the sender resumes transmission ac-
cording to available network bandwidth. Simulation results
provided in [59] indicate that TCP-Probing achieves higher
throughput rates while consuming less energy. Therefore,
the authorsbelieve that TCP-Probing provides a universal er-
ror control mechanism for heterogenous wired/wireless net-
works. The authors also present in [61] an experimental
transport protocol, called Wave and Wait Protocol (WWP),
developed specifically for a wireless environment with lim-
ited power.
8. OS/middleware and application layers
This section addresses research completed at the OS/middle-
ware and application layers with respect to energy efficiency.
8.1. OS/middleware
One important advantage of integrating wireless communi-cation with computing is that it facilitates user mobility and
connectivity to the network. Mobility, directly or indirectly,
impacts the design of operating systems, middleware, file
systems, and databases. It also presents a new set of chal-
lenges that result from power constraints and voluntary dis-
connections. To be consistent with fixed counterparts like
PCs and workstations, mobile computers need to process
multimedia information. However, such processing is ex-
pensive in terms of both bandwidth and battery power. In
general, the majority of the techniques used in the design
of todays applications to conserve bandwidth also conserve
battery life.
The main function of an operating system is to manage
access to physical resources like CPU, memory, and disk
space from the applications running on the host. To reduce
power dissipation, CPUs used in the design of portable de-
vices can be operated at lower speeds by scaling down the
supply voltage [10]. Due to the quadratic relationship be-
tween power and supply voltage, halving the supply volt-
age results in one fourth of the power being consumed. To
maintain the same throughput, the reduction in circuit speed
can be compensated by architectural techniques like pipelin-
ing and parallelism. These techniques increase throughput
resulting in an energy efficient system operating at a lower
voltage but with the same throughput. The operating systemis active in relating scheduling and delay to speed changes.
Another technique of power management at this layer is
predictive shutdown [10]. This method exploits the event
driven nature of computing in that sporadic computation ac-
tivity is triggered by external events and separated by peri-
ods of inactivity. A straightforward means of reducing aver-
age energy consumption is to shut down the system duringperiods of inactivity. However, preserving the latency and
throughput of applications requires intelligent activity-based
predictive shutdown strategies.
In [31], a study of different page placement algorithms
that exploit the new power management features of mem-
ory technology is presented. The study considers DRAM
chips that support different power modes: active, standby,
nap and powerdown. Trace-driven and execution-driven
simulations show that improvement of 6% to 55% in the
Energy Delay metric are obtained using power-aware
page allocation mechanisms that operate in conjunction with
hardware policies.
CPU scheduling techniques that attempt to minimize
power consumption are presented in [36,65]. The impact
of software architecture on power consumption is studied in
[42,58].
8.2. Application layer
The application layer in a wireless system is responsible for
such things as partitioning of tasks between the fixed and
mobile hosts, audio and video source encoding/decoding,
and context adaptation in a mobile environment. Energyeffi-
ciency at the application layer is becoming an important area
of research as is indicated by industry. APIs such as Ad-vanced Configuration and Power Interface [27] and power
management analysis tools such as Power Monitor [26] are
being developed to assist software developers in creating
programs that are more power conserving. Another power
management tool developed at Carnegie Mellon University
is PowerScope [20]. PowerScope maps energy consumption
to program structure, producing a profile of energy usage
by process and procedure. The authors report a 46% reduc-
tion in energy consumption of an adaptive video playing ap-
plication by taking advantage of the information provided
by PowerScope. This section summarizes some of the re-
search being conducted at the application layer with respect
to power conservation.
Load partitioning. Challenged by power and bandwidth
constraints, applications may be selectively partitioned be-
tween the mobile and base station [43,65]. Thus, most of the
power intensive computations of an application are executed
at the base station, and the mobile host plays the role of an
intelligent terminal for displaying and acquiring multimedia
data [43].
Proxies. Another means of managing energy and band-
width for applications on mobile clients is to use proxies.
Proxies are middleware that automatically adapt the applica-tions to changes in battery power and bandwidth. A simple
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
13/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 355
example of proxy usage during multimedia transmissions in
a low-power or low bandwidth environment is to suppress
video and permit only audio streams. Another example is to
direct a file to be printed at the nearest printer when the host
is mobile. Proxies are either on the mobile or base station
side of the wireless link.
Databases. Impact of power efficiency on database sys-
tems is considered by some researchers. For example, en-
ergy efficiency in database design by minimizing power con-
sumed per transaction through embedded indexing has been
addressed in [24]. By embedding the directory in the form of
an index, the mobile only needs to become active when data
of interest is being broadcast (the system architecture con-
sists of a single broadcast channel). When a mobile needs a
piece of information an initial probe is made into the broad-
cast channel. The mobile is then able to determine the next
occurrence of the required index and entersprobe waitmode
while it waits for the index to be broadcast. After receivingthe index information relevant to the required data, the mo-
bile enters bcast waitmode while it waits for the information
to be broadcast. Access time is defined as the sum of the two
waiting periods, probe wait and bcast wait. The goal of the
authors is to provide methods to combine index information
together with data on the single broadcast channel in order to
minimize access time. The authors propose two such strate-
gies which are further described in [24]. Also, energy effi-
cient query optimization for database systems is described
in [3].
Video processing. Multimedia processing and transmissionrequire considerable battery power as well as network band-
width. This is especially true for video processing and trans-
mission. However, reducing the effective bit rate of video
transmissions allows lightweight video encoding and decod-
ing techniques to be utilized thereby reducing power con-
sumption. Under severe bandwidth constraints or low-power
situations, video frames can even be carefully discarded be-
fore transmission while maintaining tolerable video quality.
In [1], research on processing encoded video for trans-
mission under low battery power conditions is presented.
The basic idea of this work is to decrease the number of bits
transmitted over the wireless link in response to low-power
situations. The challenge is to accomplish this goal while
preserving or minimally degrading the video quality. De-
creasing the number of transmitted bits reduces the energy
consumption due to reduced transmitter usage. In fact, sev-
eral studies have shown that transmission accounts for more
than a third of the energy consumption in video processing
and exchange in a portable device. The reduction in the num-
ber of bits can be achieved in one of two ways: (i) reducing
the number of bits in the compressed video stream generated
by the video encoder, and (ii) discarding selected packets at
the wireless network interface card (WNIC).
The first approach is possible only if two conditions are
satisfied. The portable device must be encoding a videostream as opposed to transmitting a stored video, and the ap-
plication must be able to modify parameters in the video en-
coder. The second approach is possible if the WNIC is flexi-
ble and sensitive to battery power conditions. Further details
on how the different encoding schemes affect the choice of
discarding may be found in [1]. Also, a testbed implement-
ing this research was developed, and preliminary results re-ported in [40].
Power-aware video processing is an important and excit-
ing topic. There are several approaches for developing ef-
ficient encoding schemes that will impact performance and
energy consumption as in [35,57]. However, a complete dis-
cussion is not presented due to space constraints.
9. Summary
As wireless services continue to add more capabilities such
as multimedia and QoS, low-power design remains one of
the most important research areas within wireless communi-
cation. Research must focus on decreasing the amount of en-
ergy consumed by the wireless terminal. Power conservation
has typically been considered at the physical layer. However,
most of the energy savings at the physical layer have already
been achieved. Therefore, the key to energy conservation in
wireless communications lies within the higher levels of the
wireless protocol stack. This paper describes research com-
pleted at the data link, network, transport, OS/middleware,
and application protocol layers that have addressed energy
efficiency for wireless networks. However, power conserva-
tion within the wireless protocol stack remains a very crucial
research area for the viability of wireless services in the fu-ture.
Acknowledgements
The authors wish to thank the reviewers and the Editor for
their valuable suggestions and comments that helped im-
prove the paper; and Stephanie Lindsey and Harini Krish-
namurthy who assisted with editing the paper.
References
[1] P. Agrawal, J.-C. Chen, S. Kishore, P. Ramanathan and K.M. Siva-
lingam, Battery power sensitive video processing in wireless network-
ing, in: Proc. IEEE PIMRC 98 (1998).
[2] P. Agrawal, B. Narendran, J. Sienicki and S. Yajnik, An adaptive
power control and coding scheme for mobile radio systems, in: Proc.
of IEEE Personal Communications Conf. (New Delhi, India, 1996)
pp. 283288.
[3] R. Alonso and S. Ganguly, Energy efficient query optimization, Tech-
nical report MITL-TR-33-92, Matsushita Info Tech Lab, Princeton,
NJ (1993).
[4] E. Ayanoglu, S. Paul, T.F. LaPorta, K.K. Sabnani and R.D. Gitlin,
AIRMAIL: A link-layer protocol for wireless networks, Wireless Net-
works 1(2) (1995) 161174.
[5] A. Bakre and B.R. Badrinath, I-TCP: indirect TCP for mobile hosts,
in: Proc. of 15th Int. Conf. on Distributed Computing Systems (1995)
pp. 136143.
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
14/16
356 JONES ET AL.
[6] H. Balakrishnan, S. Seshan and R. Katz, Improving reliable transport
and handoff performance in cellular wireless networks, Wireless Net-
works 1(4) (1995).
[7] N. Bambos, Toward power-sensitive network architectures in wireless
communications: concepts, issues and design aspects, IEEE Personal
Communications 5(3) (1998) 5059.
[8] K. Brown and S. Singh, M-TCP: TCP for mobile cellular networks,Computer Communication Review 27(5) (1997) 1943.
[9] R. Cceres and L. Iftode, Improving the performance of reliable trans-
port protocols in mobile computing environments, IEEE Journal on
Selected Areas in Communications 13(5) (1995) 850857.
[10] A. Chandrakasan and R.W. Brodersen, Low Power Digital CMOS De-
sign (Kluwer Academic, Norwell, MA, 1995).
[11] J.-H. Chang and L. Tassiulas, Energy conserving routing in wireless
ad-hoc networks, in: Proc. IEEE INFOCOM, Tel-Aviv, Israel (2000).
[12] J.-C. Chen, K.M. Sivalingam and P. Agrawal, Performance compari-
son of battery power consumption in wireless multiple access proto-
cols, Wireless Networks 5(6) (1999) 445460.
[13] J.-C. Chen, K.M. Sivalingam, P. Agrawal and R. Acharya, Scheduling
multimedia services for a low-power MAC in wireless and mobile
ATM networks, IEEE Transactions on Multimedia 1(2) (1999) 187
201.[14] L. Chen, Mobile-based error compensation algorithms for wireless
networks, Masters thesis, Washington State University, Pullman
(2000).
[15] S. Damodaran and K.M. Sivalingam, Adaptive scheduling at mobiles
for wireless networks with multiple priority traffic and multiple trans-
mission channels, in: Proc. of Intl. Conf. on High Performance Com-
puting, Calcutta, India (1999).
[16] I.B. Dhaou, A novel load-sharing algorithm for energy efficient MAC
protocol compliant with 802.11 WLAN, in: Proc. IEEE Vehicular
Technology Conference, Amsterdam, Holland (1999).
[17] F. Douglis, P. Krishnan and B. Marsh, Thwarting the power-hungry
disk, in: Proceedings of the 1994 Winter USENIX Conference (1994).
[18] L.M. Feeney, An energy consumption model for performance analysis
of routing protocols for mobile ad hoc networks, in: Proc. of the 45th
IETF Meeting: MANET Working Group (1999).[19] L.M. Feeney, Investigating the energy consumption of an IEEE
802.11 network interface, Technical report T1999-11, Swedish Insti-
tute of Computer Science, Kista, Sweden (1999).
[20] J. Flinn and M. Satyanarayanan, PowerScope: A tool for profiling the
energy usage of mobile applications, in: Proceedings of the 2nd IEEE
Workshop on Mobile Computing Systems and Applications (1999),
http://www.cs.cmu.edu/user/jflinn/www/pscope.html.
[21] B.M. Gordon, E. Tsern and T.H. Meng, Design of a low power video
decompression chip set for portable applications, Journal of VLSI
Signal Processing Systems 13 (1996) 125142.
[22] K. Govil, E. Chan and H. Wasserman, Comparing algorithms for dy-
namic speed-setting of a low-power CPU, in: Proc. Mobicom. (1995)
pp. 1325.
[23] IEEE, Wireless LAN medium access control (MAC) and physical
layer (PHY) Spec, IEEE 802.11 standard (1998).[24] T. Imielinski, S. Vishwanathan and B.R. Badrinath, E nergy efficient
indexing on air, in: Proceedings of the International Conference on
Management of Data (ACM-SIGMOD) (1994).
[25] Infrared Data Association, IrDA SIR data specification (2000), http:
//www.irda.org/standards/specifications.asp.
[26] Intel Corporation, Intel Power Measurement Tools (2000), http://
developer.intel.com/design/mobile/intelpower/tools/.
[27] Intel Corporation, Microsoft and Toshiba Corporation, Advanced
Configuration & Power Interface (2000), http://www.teleport.
com/acpi/.
[28] D.B. Johnson, D.A. Maltz, Y.-C. Hu and J. G. Jetcheva, The dy-
namic source routing protocol for mobile ad hoc networks, IETF Draft
MANET Working Group (2000).
[29] P. Karn, MACA A new channel access method for packet radio, in:
ARRL/CRRL Amateur Radio 9th Computer Networking Conference(1990) pp. 134140.
[30] S. Kishore, J.-C. Chen, K.M. Sivalingam and P. Agrawal, Adaptive
power control and scheduling algorithms based on battery power level
for CDMA wireless networks, in: Proc. of IEEE ICUPC, Florence,
Italy (1998) pp. 967971.
[31] A.R. Lebeck, X. Fan, H. Zeng and C. Ellis, Power-aware page alloca-
tion, in: Proc. of Ninth ACM ASPLOS (2000) pp. 105116.
[32] P. Lettieri, C. Fragouli and M.B. Srivastava, Low power error con-trol for wireless links, in: Proc. ACM MobiCom, Budapest, Hungary
(1997).
[33] P. Lettieri and M. Srivastava, Advances in wireless terminals, IEEE
Personal Communications 6(1) (1999) 618.
[34] P. Lettieri and M. B. Srivastava, Adaptive frame length control for
improving wireless link throughput, range, and energy efficiency, in:
INFOCOM, Vol. 2 (1998) pp. 564571.
[35] H. Liu and M.E. Zarki, Adaptive source rate control for real-time
wireless video transmission, Mobile Networks and Applications 3
(1998) 4960.
[36] J. Lorch and A.J. Smith, Scheduling techniques for reducing proces-
sor energy use in MacOS, Wireless Networks 3(5) (1997) 311324.
[37] J.R. Lorch and A.J. Smith, Software strategies for portable computer
energy management, IEEE Personal Communications 5(3) (1998)
6073.[38] J. Macker and M. Corson, IETF working group: mobile ad-hoc net-
works (MANET), http://www.ietf.org/html.charters/manet-charter.
html.
[39] J. Macker and M. Corson, Mobile ad-hoc networking and the IETF,
ACM Mobile Computing and Communications Review 2(1) (1998).
[40] I. Mahadevan, Architecture and algorithms for Quality of Service
support and energy efficient protocols for wireless/mobile networks,
Ph.D. thesis, Washington State University (1999).
[41] B. Marsh, F. Douglis and P. Krishnan, Flash memory file caching for
mobile computers, in: Proceedings of the 27th Hawaii Conference on
Systems Science (1994).
[42] H. Mehta, R.M. Owens, M.J. Irwin, R. Chen and D. Ghosh,
Techniques for low energy software, in: Proc. of the International
Symposium on Low Power Electronics and Design (1997) pp. 7275.
[43] S. Narayanaswamy, S. Seshan, E. Amir, E. Brewer, R.W. Brodersen,F. Burghart, A. Burstein, Y. Chang, A. Fox, J.M. Gilbert, R. Han,
R.H. Katz, A.C. Long, D.G. Messerschmitt and J.M. Rabaey, A
low-power, lightweight unit to provide ubiquitous information
access applications and network support for Infopad, IEEE Personal
Communications (1996) 417.
[44] B. Narendran, J. Sienicki, S. Yajnik and P. Agrawal, Evaluation of
an adaptive power and error control algorithm for wireless systems,
in: Proc. of IEEE Communications Conf. Montreal, Canada (1997)
pp. 349355.
[45] C.E. Perkins, E.M. Royer and S.R. Das, Ad hoc on-demand distance
vector (AODV) routing, IETF Draft MANET Working Group (2000).
[46] J. Postel, Transmission control protocol, RFC 793 (1981).
[47] C.E. Price, Power-aware scheduling algorithms for wireless networks,
Masters thesis, Washington State University, Pullman (2000).
[48] C.E. Price, K.M. Sivalingam, J.-C. Chen and P. Agrawal, Power-awarescheduling algorithms for wireless networks, in: Proc. Intl. Confer-
ence on Intelligence Computing and VLSI, Kalyani, India (2001).
[49] R. Ramanathan and R. Rosales-Hain, Topology Control of multihop
wireless networks using transmit power adjustment, in: Proc. IEEE
INFOCOM, Tel-Aviv, Israel (2000).
[50] A. Salkintzis and P.T. Mathiopoulos (Guest Eds.), The Evolution
of Mobile Data Networking, IEEE Personal Communications 3(2)
(2000).
[51] S. Singh and C.S. Raghavendra, PAMAS: Power aware multi-
access protocol with signalling for ad hoc networks, Computer
Communication Review 28(3) (1998) 526.
[52] S. Singh, C.S. Raghavendra and J. Stephanek, Power-aware broad-
casting in mobile ad hoc networks, Technical report, Oregon State
University, Department of Electrical and Computer Engineering
(1999).[53] K.M. Sivalingam, J.-C. Chen, P. Agrawal and M. Srivastava, Design
-
7/27/2019 A Survey of Energy Efficient Network Protocols ForWireless Networks
15/16
A SURVEY OF ENERGY EFFICIENT NETWORK PROTOCOLS 357
and analysis of low-power access protocols for wireless and mobile
ATM networks, Wireless Networks 6(1) (2000) 7387.
[54] M. Srivastava, TUTORIAL: energy efficiency in mobile computing
and networking, in: ACM MobiCom Tutorials, Boston, MA (2000).
[55] M. Stemm and R.H. Katz, Measuring and reducing energy consump-
tion of network interfaces in hand-held devices, IEICE Transactions
on Fundamentals of Electronics, Communications, and ComputerScience (1997).
[56] I. Stojmenovic and X. Lin, Power-aware localized routing in wireless
networks, in: Proc. of the IEEE International Parallel and Distributed
Processing Symposium, Cancum, Mexico (2000).
[57] R. Swann and N. Kingsbury, Bandwidth efficient transmission of
MPEG-II video over noisy mobile links, Signal Processing 12(2)
(1998) 105115.
[58] V. Tiwari, S. Malik and A. Wolfe, Power analysis of embedded
software: A first step towards software power minimization, IEEE
Transactions on Very Large Scale Integration 2(4) (1994) 437445.
[59] V. Tsaoussidis and H. Badr, TCP-probing: towards an error control
schema with energy and throughput performance gain, in: Interna-
tional Conference on Network Protocols, Osaka, Japan (2000).
[60] V. Tsaoussidis, H. Badr, X. Ge and K. Pentikousis, Energy/throughput
tradeoffs of TCP error control strategies, in: Proc. of 5th IEEE Sym-posium on Computers and Communications (ISCC) 2000, France
(2000).
[61] V. Tsaoussidis, A. Lahanas and H. Badr, Wave & Wait protocol
(WWP): low energy, high throughput for mobile IP-devices, in: Proc.
of 8th IEEE Conference on Networks (ICON) 2000, Singapore (2000).
[62] S. Udani and J. Smith, Power management in mobile computing (a
survey), University of Pennsylvania (1996), http://www.cis.upenn.
edu/udani/papers.html.
[63] N. Vaidya, Tutorial: Mobile ad hoc networks: routing, MAC and
transport issues, in: ACM MobiCom Tutorials, Boston, MA (2000).
[64] WAP Forum, Wireless application protocol (WAP) architecture
specification, http://www.wapforum.org/what/technical.htm (2000).
[65] M. Weiser, B. Welch, A. Demers and S. Shenker, Scheduling for
Reduced CPU Energy, in: Proc. of First Symposium on Operating
Systems Design and Implementation (OSDI), USENIX Association,Monterey, CA (1994).
[66] J.E. Wieselthier, G.D. Nguyen and A. Ephremides, On the construc-
tion and energy-efficient broadcast and multicast trees in wireless
networks, in: Proc. IEEE INFOCOM, Tel-Aviv, Israel (2000).
[67] H. Woesner, J.-P. Ebert, M. Schlager and A. Wolisz, Power-saving
mechanisms in emerging standards for wireless LANs: the MAC
level perspective, IEEE Personal Communications 5(3) (1998) 4048.
[68] M. Woo, S. Singh and C.S. Raghavendra, Power aware routing in
mobile ad hoc networks, in: Proc. ACM MobiCom, Dallas, TX (1998)
pp. 181190.
[69] M. Zorzi and R. Rao, Energy constrained error control for wireless
channels, IEEE Personal Communications (1997).
[70] M. Zorzi and R.R. Rao, Error control and energy consumption in
communications for nomadic computing, IEEE Transactions on
Computers 46(3) (1997) 279289.[71] M. Zorzi and R.R. Rao, Energy efficiency of TCP in a local wireless
environment, Mobile Networks and Applications (2000) accepted.
[72] M. Zorzi (Guest Ed.), Energy Management in Personal Communica-
tions and Mobile Computing, IEEE Personal Communications 5(3)
(1998).
Christine Jones received her M.S. in computer sci-
ence from Washington State University, Pullman in
2000 and her B.S. in computer science from Whit-
worth College in 1998. She is presently with BBN
Technologies in Cambridge, MA, USA.
E-mail: [email protected]
Krishna M. Sivalingam (ACM 93, IEEE SM 00
M 95) received his Ph.D. and M.S. degrees in com-
puter science from State University of New York at
Buffalo in 1994 and 1990, respectively. While at
SUNY Buffalo, he was a Presidential Fellow from
1988 to 1991. Prior to that, he received the B.E. de-
gree in computer science and engineering in 1988
from Anna University, Madras, India. He is Boe-
ing Associate Professor of Computer Science inthe School of Electrical Engineering and Computer
Science, at Washington State University, Pullman, where he was an Assis-
tant Professor from 1997 to 2000. Earlier, he was an Assistant Professor
at University of North Carolina Greensboro from 1994 until 1997. He has
conducted research at Lucent Technologies Bell Labs in Murray Hill, NJ,
and at AT&T Labs in Whippany, NJ.
His research interests include wireless networks, optical wavelength di-
vision multiplexed networks, and performance evaluation. He has served as
a Guest Co-Editor for a special issue of the IEEE Journal on Selected Areas
in Communications on